Last updated:
To check GPU health, you monitor the graphics card’s temperature, clock speed, fan speed and power under load, run a stress test for 15–30 minutes, and watch for crashes, artifacts or throttling. A healthy GPU stays stable, stays below its rated maximum temperature and holds its boost clock.
What are the signs of a healthy GPU?
A healthy GPU shows 5 signs:
- Stable temperatures within the normal GPU temp range.
- Consistent boost clock under load without sudden drops.
- Spinning fans that respond to temperature.
- No artifacts, crashes or driver resets.
- Expected performance compared with benchmark results for the same model.
What are the signs of a failing GPU?
A failing GPU shows 6 warning signs:
- Artifacts: dots, lines or flickering; see GPU artifacting.
- Crashes and driver timeouts in games.
- Black screens or no display.
- Overheating even after cleaning.
- Loud, grinding or stopped fans.
- Performance far below the expected level.
How do you check GPU health step by step?
You check GPU health with these 7 tests:
- Check Device Manager. Open Device Manager > Display adapters. A yellow warning icon means a driver or hardware error.
- Read the sensors. Open HWiNFO or GPU-Z and note idle temperature, clocks and fan speed; see how to check GPU temp.
- Run a stress test. Run 3DMark Time Spy stress test, Unigine Superposition or OCCT 3D for 15–30 minutes.
- Watch temperature and clocks. Temperature should level off below the maximum and clocks should stay steady.
- Test the VRAM. Run OCCT VRAM test or MemTestCL/MemtestG80 to find memory errors.
- Compare a benchmark score. Compare your 3DMark score with the average score for your GPU model; a result 10%+ lower points to throttling or a fault.
- Check the event log. In Event Viewer, search for nvlddmkm or amdkmdag errors, which mean driver timeouts; see error occurred on GPUID 100.
How long should you stress test a GPU?
You should stress test a GPU for 15–30 minutes for a health check and 1 hour when buying a used card or testing an overclock. A 3DMark stress test passes with 97% frame rate stability or higher.
How do you test a GPU without a game?
You test a GPU without a game with free benchmark tools: 3DMark (demo version on Steam), Unigine Superposition, OCCT and FurMark. FurMark creates extreme heat, so use it for short checks only.
How do you check the health of a used GPU before buying?
You check the health of a used GPU by running a stress test, a VRAM test and a benchmark in person, checking the fans and the PCIe connector, and inspecting the card for burnt marks or missing screws. The full checklist is in buying a used GPU.
How long does a GPU last?
A GPU lasts 5–8 years on average. Fans and thermal paste often wear out first and are cheap to replace.
FAQ
Can I check GPU health without software?
Partly. Device Manager and Task Manager show errors and temperature, but a stress test needs software.
Is FurMark safe for my GPU?
FurMark is safe for short tests on healthy cards, but it creates unrealistic heat. 3DMark and OCCT are better for long tests.
What is a good 3DMark stress test result?
A good 3DMark stress test result is 97% frame rate stability or higher.
Related guides
If your card is getting old, our guide on how long GPUs last explains which parts wear out first and how to extend its life.
2 thoughts on “How to Check GPU Health: 7 Tests for Your Graphics Card”